Decoding Hospital Evacuation Drills: Pre-movement and Movement Analysis in New Zealand

Fire and evacuation modelling tools are becoming increasingly popular in fire engineering to carry out performance-based assessments of the built environment. The latest survey on the use of computer-based evacuation modeling tools shows that users can choose among over 70 options [1]. Notwithstanding the availability of simulation tools, one of the main challenges for fire engineers is the selection of appropriate input values. These input values are fundamental to defining the pre-evacuation and movement behaviors of the simulated evacuees and can significantly impact the final output of the model [2‐4]. The selection of reasonable values is one of the most important tasks while using existing models. Fire engineers are required to make reasonable assumptions and to identify suitable values from existing behavioral studies or reports. As such, evacuation databases represent one of the most useful tools to support fire engineering practitioners in this challenging task.

The pioneering work of Fahy and Proulx [5] provides the first instance of an evacuation database in the fire engineering field. In fact, they acknowledged and foresaw the need for lists of existing evacuation studies providing quantitative data on pre-evacuation times, walking speeds, and evacuees’ characteristics and actions. As such, Fahy and Proulx established a strategic pathway for the field. Today, many new evacuation studies have been published, providing new evacuation data that is usable to improve assumptions on evacuation simulations. Following the pathway of Fahy and Proulx [5], new databases have been proposed in the last 20 years by Shi et al. [6], Gwynne and Boyce [7], Geoerg et al. [8] and Lovreglio et al. [9]. Although these databases aim to provide a comprehensive overview of all the types of occupancy, there is still a clear gap on some specific types of occupancies requiring assisted evacuation, such as hospitals, healthcare facilities, and retirement homes. A partial answer to this need is the database by Geoerg et al. [8] who provide data focusing on evacuees with reduced mobilities. However, this work does not include a comprehensive analysis of hospital evacuations.

Simulating hospital evacuation is still a major challenge. This task can be done assuming both a total and staged evacuation depending on the fire safety strategy and building design aspects (e.g., fire compartments, fire safety systems, and evacuation zones). Today, different evacuation tools have been adapted to simulate assisted evacuations for hospitals [10‐12]. Most of these tools allow engineers to simulate hospital staff to assist patients with mobility impairments evacuating them using different devices such as beds, chairs, or other specialized devices [12]. In this type of simulation, fire engineers are often asked to make sensitive choices when setting up the hospital simulations. However, this task is still very challenging given the lack of an evacuation database on preparation times and evacuation speeds using hospital evacuation equipment. This is also partially due to the lack of existing data from hospital evacuations.

This work aims to provide new instruments for fire engineers to perform reliable evacuation simulations for hospitals. To achieve this target, we provide new evacuation data collected from eight evacuation exercises carried out in two hospitals in New Zealand. These exercises were video recorded by the research team and the video footage was analyzed to generate new data on assisted evacuation times and speeds in hospitals. Further, we generate a new database combining this new data on preparation times, pre-movement times, and evacuation speeds with the existing data from the existing evacuation literature.

Figure 10 summarises the pre-movement times observed in various empirical studies related to healthcare facilities. Generally, most studies report pre-movement times within a similar range of approximately 50–300 s. Additionally, the study by Gwynne et al. [13] highlights significant differences in pre-movement times between a hospital outpatient area and a university library, emphasizing the dependency on staff instructions, which can delay evacuation initiation in hospitals.

This indicates a general consistency in response times across different healthcare settings and populations. However, it is noteworthy that some drills exhibit outliers and extremes, as reported by Schliephake [14] and Kuligowski et al. [15], showing significantly higher pre-movement times extending beyond 500 s. These observations suggest the presence of unobserved heterogeneity, such as the influence of specific population characteristics or the boundaries of facilities, resulting in prolonged initiation of the evacuation process.

Focusing on healthcare facilities, three general findings are striking:

The data on the movement speeds of evacuees in healthcare contexts highlights significant variability, which is largely influenced by the type of assistive devices used, the physical capabilities of the individuals, and the specific evacuation scenarios (cf. Figure 11 for comparison). On average, movement speeds range from approximately 0.04 to 1.54 \({m}\, {s}^{-1}\).

Individuals with higher mobility levels or those using advanced assistive devices exhibit faster movement speeds. For example, Geoerg et al. [21] found that participants without disabilities demonstrated lower movement times (resulting in higher speeds) compared to those with disabilities; the level of assistance required also impacted movement speeds. In the study by Hunt et al. [20], devices requiring less manual effort from handlers (e.g., evacuation chairs) allowed for quicker movement compared to those requiring significant physical exertion (e.g., stretchers and rescue sheets).

The effectiveness of evacuation is enhanced with trained staff, as highlighted by Strating [18]. Hunt et al. [20] emphasized that trained hospital staff using evacuation chairs achieved consistent and efficient speeds, demonstrating the importance of regular training and drills. Unannounced drills, as conducted by Geoerg et al. [21], revealed higher variability in movement speeds, which can be attributed to the element of surprise and the real-time decision-making required during such evacuations.

Our study aims to bridge these gaps by focusing specifically on hospitals rather than general healthcare facilities. We emphasize the importance of conducting research under ’realistic’ conditions in natural environments, using actual hospital staff instead of actors.

Additionally, it is crucial to consider the different occupancy types within a hospital to gain a comprehensive understanding of evacuation dynamics. Previous studies suggest that various assistive devices, such as wheelchairs, beds, and other mobility aids, significantly influence evacuation outcomes. However, these factors have not been adequately explored. By including these elements in our investigation, we aim to accurately reflect the diverse needs and challenges of hospital evacuations.

In 2020 a series of eight announced evacuation drills were conducted across two hospitals in New Zealand (Hospital A and Hospital B).

The evacuation drills were carried out across several occupancy types within the hospitals.

This section describes the results obtained from the evacuation drills, beginning with qualitative observations before exploring the quantitative results.

This work provides new insights to aid fire engineers in performing reliable evacuation simulations for hospitals. This was done by collecting new evacuation data from eight evacuation exercises carried out in two hospitals in New Zealand. The data was collected by video recording these evacuation exercises and manually extracting data on preparation times, pre-movement times, and evacuation speed. Further, the proposed data are compared and combined in this section (see Figs. 10, 11 and Table 5) to generate a new database for hospital evacuation. As such, this work is an extension of the pioneering work of Fahy and Proulx [5], who provided one of the first instances of a cohesive database of pre-movement times and walking speeds for use in evacuation modelling. The proposed finding in Sect. 4.2 shows substantial differences in the active preparation times depending on the type of area and evacuation mode (see Fig. 3). On the other hand, results in Sect. 4.3 show that the passive preparation times were equivalent when comparing the data for evacuation mode. However, there were substantial differences when aggregating the data by type of area (see Fig. 5). These findings suggest that it is necessary to consider personalized evacuation strategies that account for the different evacuation needs when running hospital evacuations. In fact, both evacuees with impairment and bedridden evacuees require more tailored solutions, which include better training for staff and the use of effective assistive devices. When comparing the passive and active times in Fig. 2, it is possible to observe that, except for HDU evacuations, the passive preparation times are substantially greater than the active preparation times. Noting that this passive phase is when there is no direct action toward effective evacuation, this indicates that the pre-evacuation time can be dramatically reduced, by increasing both the ratio of patients/staff and reducing the waiting time for equipment. Another interesting finding from the proposed evacuation investigation is the non-sequential behaviour of staff preparation activities. In fact, in several instances, staff did not complete the preparation activities for a single patient in a single time period. In fact, when staff were responsible for multiple patients, they did not prepare each one fully in sequence, but they were passing from one patient to another. When comparing the proposed pre-movement times with the one proposed in the literature, the results in Fig. 10 show that this new data has substantially higher average values than many other existing studies. This could be due to the evacuation drills in this study being conducted in working hospital wards with many patients, when compared with some of the existing data sources which were in outpatients facilities [13], office or residential buildings [15, 22‐24], or used simplified and well-defined scenarios for research purposes with high staff:patient ratios [18, 19, 21, 25]. On the other hand, one of the limitations of this work is that it does not provide insight into the time required for specific activities during the active preparation times. This is due to the type of evacuation exercises which were not designed for this purpose. When focusing on the horizontal evacuation speed in Sect. 4.5, the results show that bed evacuation speeds are lower than walking and wheelchair evacuation speeds (see Fig. 7). On the other hand, the data in Fig. 2 highlights that the travel time is a relatively small component of the total evacuation with only a few outliers. This is because congestion was not observed in the evacuation exercises. As such, this work indicates the need to put more focus on the pre-evacuation stage when planning evacuations and/or running simulations. Finally, the proposed horizontal speed clearly overlaps with the data reported in the literature, as shown in Fig. 11.

With any experiment, there is a balance to be made between external and internal validity [26, 27]. Announced evacuation drills were the method used in these drills, primarily due to ethical limitations around conducting unannounced evacuation drills on real patients within each of these hospital spaces. While there is a reduction in urgency and surprise associated with an announced evacuation, there are still meaningful insights to be gained with such an activity. Further, only a single sensory input (aural alerting devices) was provided to participants. The use of other sensory inputs, such as visual, olfactory, or somatosensory inputs, is likely to improve a sense of realism and motivation and would be useful to explore in future work. Patients were healthy volunteers and not actual patients within the hospital, this was due to ethical considerations and to work within the operational requirements of a working hospital. The use of healthy volunteers aids in the reduction in realism for staff, having patients who are not only dressed differently and lying on top of a bed but who are not physically connected to the medical devices within the hospital. One avenue for future exploration could be to use actors who are dressed like patients, lying in bed, and actually connected (where feasible) to monitoring and medical equipment. This scenario could be further enhanced by limiting the number of observers who both clutter the area of observation and also add to the artifice of the scenario, as well as adding a narrative portion to describe the situation as not just a drill but as an actual emergency evacuation. Finally, there were observable differences between types of spaces within the hospital evacuation drills carried out in this study, particularly between higher-acuity spaces and low-acuity spaces (e.g. GW spaces). There is a need to gain a better understanding of some of the differences between high-acuity spaces (e.g. HDU) and low-acuity spaces (e.g. GW spaces). It seems reasonable to deduce this difference is due to additional medical equipment, which takes more time to disconnect. However, a detailed simulation study would be beneficial in collecting this type of data on the relationship between medical acuity and preparation times.

Bild vergrößern

Bild vergrößern

The findings from this research highlight some of the nuanced behaviours which occur during the pre-movement phases of the assisted evacuation of inpatients from a hospital facility. An understanding of the presence and magnitude of these components is crucial to enable appropriate inputs for modelling hospital spaces where patients are unable to egress without assistance.

Results suggest that depending on the type of space, there can be considerable differences in the proportion of time spent actively preparing a patient for egress versus time spent waiting for equipment or staff to become available. This finding is important as it highlights the influence of staff:patient ratios and the availability of required portable medical equipment, on the assisted evacuation of inpatients.

For the highest acuity spaces studied (HDU), it was observed that the high staff:patient ratios meant that the passive preparation times were very minimal with a high proportion of the time spent actively preparing the patient for egress. Even with this considerable resource available these patients represented the longest pre-movement times by a large margin. The reason for this is the additional equipment used for these patients which required long preparation times to disconnect and make ready for evacuation.

Overall, this research highlights the value to be gained by a deeper understanding of the sub-components that comprise the pre-movement phases of evacuation in a hospital occupancy. While conducted in hospitals, the results have value for other occupancy types where evacuees require assistance to evacuate. While this work contributes useful data for egress modelling, it also paves the way for further work to develop an understanding of the factors which influence the active preparation phases such as the level of acuity and medical equipment.